Big Idea

The argument in one line.

Claude Code released ten hidden features over the past month—session resuming, context visualization, custom memory, YOLO mode, and a plugin store—that fundamentally change how developers should structure their workflow and manage AI collaboration.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You're actively using Claude Code to build projects and want to learn 10 newly-shipped features most people haven't discovered yet.
A developer who's frustrated with Claude Code performance degradation and needs concrete strategies to manage context window bloat.
You use Claude Code regularly but haven't explored its terminal commands and advanced session management tools.

SKIP IF…

You don't use Claude Code at all or are still deciding whether to adopt it — this is entirely feature-focused.
You're looking for foundational Claude Code setup or workflow design — this assumes you're already building with it.

TL;DR

The full version, fast.

Claude Code shipped ten quiet updates over the holidays that change how you actually drive it, and most users never noticed. The core mechanic is treating Claude Code less like a chat window and more like a managed workspace: resume past sessions with claude --resume, name them with /rename, audit your context window with /context, monitor plan usage with /stats, and rewind mistakes by double-tapping escape instead of starting over. Force harder reasoning on complex tasks by typing ultrathink in the prompt, pin behaviors with the hashtag memory shortcut instead of bloating CLAUDE.md, stash drafts with Ctrl+S, and install skills like front-end design from /plugins. Run long unattended jobs with claude --dangerously-skip-permissions only once your prompting is dialed in.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 00:45

01 · Hook + Promise

FOMO open: 10 secret updates, explicit beginner-to-pro promise, description CTA established

00:45 – 01:58

02 · 1. Resume Sessions

claude --resume opens a picker of all past sessions for the project; demo on a Kanban project management app

01:58 – 03:18

03 · 2. Visualize Context (/context)

/context shows a breakdown of what is filling the context window (system prompt, tools, messages); /clear to free up the messages portion

03:18 – 04:17

04 · 3. Usage Stats (/stats)

/stats shows token usage, streaks, active days, peak hours, and model breakdown; community engagement hook: share your streak in comments

04:17 – 05:44

05 · 4. Name Sessions (/rename)

/rename <name> tags the session so --resume picker shows meaningful names; demo: calling a session Kanban

05:44 – 07:14

06 · 5. Rewind (Double Escape)

Double-press Escape triggers a rewind picker showing all recent checkpoints; select one to restore code and conversation state

07:14 – 08:55

07 · 6. UltraThink

Add the word Ultrathink to any prompt to trigger extended thinking; renders in rainbow colors in the terminal; recommended for complex tasks and initial build phases

08:55 – 10:47

08 · 7. Custom Memories (#)

Type # to enter memory mode; specify user memory or project memory; recommended homework: always use the design skill when changing the UI

10:47 – 11:39

09 · 8. YOLO Mode (--dangerously-skip-permissions)

claude --dangerously-skip-permissions disables permission prompts; recommended only for experienced users; enables long-running autonomous tasks

11:39 – 13:00

10 · 9. Plugin Store (/plugins)

/plugins opens a discoverable store of skills, plugins, and MCPs; top pick: frontend design skill

13:00 – 13:40

11 · 10. Stash Prompts (Ctrl+S)

Ctrl+S stashes the current prompt draft; after a quick command the stash auto-restores to the input line

13:40 – 14:27

12 · Outro + CTA

Subscribe, notifications, like; livestreams Mon/Wed/Fri 11AM PST; engagement question: which update will you use?

Atomic Insights

Lines worth screenshotting.

Claude --resume surfaces every past session for a project so you can pick up exactly where you left off without re-explaining context to the model.
The /context slash command visualizes what is consuming your context window by category — system prompt, tools, and message history — so you can act on the right fix.
Message history is typically the largest context consumer; /clear removes it while leaving the system prompt and tools intact.
/stats shows token usage, active streaks, session counts, and model breakdown so you can monitor your plan consumption before hitting a limit.
/rename gives a session a human-readable name so that when you resume from a list of sessions, you know instantly which conversation to pick.
Running multiple Claude Code sessions in parallel with distinct names enables parallel feature development without context cross-contamination.
Double-escape triggers a rewind menu that lets you restore code and conversation to any prior checkpoint without a manual git reset.
YOLO mode bypasses all confirmation prompts so Claude Code executes every command autonomously — appropriate for trusted workflows, not experimental ones.
A plugin store for Claude Code means the tool's capability surface is no longer limited to what Anthropic ships — third-party integrations can extend it.
Session naming is a prerequisite for the resume feature to be useful at scale — unnamed sessions become indistinguishable in a long history list.
Context visualization turns a vague 'Claude is getting worse' complaint into a specific, addressable root cause — usually message history bloat.
The combination of resume, rename, and context commands creates a session management discipline that compounds over time as project complexity grows.

Takeaway

Ten Claude Code Features Most Users Have Never Touched

Claude Code updates

Alex Finn surfaces ten quietly shipped Claude Code features — session recovery, context visualization, rewind, YOLO mode, and the plugin store — that most users missed and that change the daily workflow.

021. Resume Sessions

claude --resume opens a picker of all past sessions for the project — context is no longer tied to the terminal window
Every previous session with its full context becomes a resumable starting point

032. Visualize Context (/context)

/context shows the breakdown of what is filling the context window: system prompt, tools, messages
/clear frees up only the messages portion without resetting settings or tools

043. Usage Stats (/stats)

/stats shows token usage, streaks, active days, peak hours, and model breakdown
The streak data creates a consistency feedback loop — the same mechanism that makes habit trackers work

054. Name Sessions (/rename)

/rename before you close a session tags it with a meaningful name in the resume picker
Named sessions make it easy to return to the right context without guessing from timestamps

065. Rewind (Double Escape)

Double-press Escape to open a rewind picker showing recent checkpoints
Selecting a checkpoint restores both code state and conversation state — full undo for an AI coding session

076. UltraThink

Add Ultrathink to any prompt to trigger extended thinking — renders in rainbow colors in the terminal
Use it for complex architectural decisions and initial build phases, not routine edits where standard thinking is sufficient

087. Custom Memories (#)

Type # to enter memory mode and add a persistent rule at user or project level
Project memories like always use the design skill when changing the UI apply in every future session without re-explanation

098. YOLO Mode (--dangerously-skip-permissions)

Disables permission prompts for long-running autonomous tasks — use only when you trust the session completely
YOLO mode is the unlock for overnight agent runs that would otherwise stall waiting for approvals

109. Plugin Store (/plugins)

/plugins opens a discoverable marketplace of skills, plugins, and MCPs directly from the terminal
Frontend design skill is the top recommended first install from the store

1110. Stash Prompts (Ctrl+S)

Ctrl+S stashes the current prompt draft before running a quick command
The stash auto-restores when you return to the input line — no more losing a long prompt to a fast command

Glossary

Terms worth knowing.

Context window: The maximum amount of text (previous messages, instructions, code, files) that an AI model can hold in active memory at once; exceeding it causes the model to lose track of earlier information.
Session resume: A Claude Code feature (--resume flag) that reloads a previous conversation including all its context, so work can continue without re-explaining the project from scratch.
YOLO mode: An optional Claude Code setting that automatically approves tool-use actions without prompting the user for confirmation, allowing the AI to execute commands uninterrupted.
Slash command: A text shortcut prefixed with '/' typed in a terminal or chat interface that triggers a specific function — for example, /context to display memory usage or /clear to wipe conversation history.
Hallucination: When an AI model generates confident-sounding but factually incorrect or fabricated information, often triggered by insufficient or overloaded context.
Claude rules: Custom instructions stored in a project's CLAUDE.md file that shape how Claude Code behaves — defining coding style, allowed actions, project conventions, and other persistent preferences.
Plugin store (Claude Code): A marketplace of extensions and integrations for Claude Code that add new capabilities, tools, or workflows beyond the default feature set.
Context visualization: A diagnostic view (triggered by the /context command in Claude Code) showing a breakdown of what is consuming the available context window — system prompts, messages, tool outputs — so users can identify and clear the largest items.

Quotables

Lines you could clip.

02:00

“I hear a lot of people on X complaining right now, oh, Claude's getting worse. No. That's a skill issue. They're just not managing their context well.”

Hot take with a direct callout — triggers algorithmic engagement and comments from both sides→ TikTok hook↗ Tweet quote

10:47

“Claude for me hasn't made mistakes in months. And I know that might sound crazy...”

Confident, polarizing claim that begs a response — strong standalone reel opener→ IG reel cold open↗ Tweet quote

07:14

“UltraThink is totally back into Claude code, and you need to be using it.”

Short, punchy, specific — works as a clip title and reel hook→ TikTok hook↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

00:00Over the past month, Anthropic has released a ton of secret updates to Claude Code that you would only find out about if you were following specific Twitter accounts. In this video, I'm gonna go over all those secret updates that you probably missed that will completely change how you use Claude Code and make it way better.

00:18If you use Claude Code, this is gonna be the most important video you watch this month. I promise you, whether you're a beginner or a pro, you will be using Claude Code completely differently after these 10 updates. Let's get into it.

00:31So this one is a great one, and that is the ability to resume past Claude code sessions. If you're anything like me, you've had tons of conversations with Claude code that has context you don't wanna miss out on. You reset your computer and all that context disappears.

00:45Well, no more. If you type into your terminal claud dash dash resume, hit enter, it will bring up all your past sessions you had with claud code for the project you're working on.

00:58So for the sake of demonstration of this video, I'm gonna show you this project management tool I'm building out right now. It helps me manage all my projects in Claude code. So I have all my past sessions here that I've had with Claude.

01:10What So I'm gonna do is I'm gonna hit resume on my past session, and all the context I had and all the conversation I had from that session is now completely resumed. So now I don't have to go back and fill in Claude code with all the past context every time I reset my computer or my Visual Studio code. This is massive, and this is gonna make it so much easier to get back into past conversations you have and not miss out on any of the context you had before.

01:36I put the command to resume your sessions down below in the description. All the commands and all the changes I go over will be down in the description below, so feel free to pause and copy and paste that into, like, your Apple notes. Okay.

01:47So say we resumed our past session and there's a ton of context in there. When you get a lot of context into Claude code and starts filling up your window, it actually can make Claude hallucinate a lot more than it does before.

02:00I hear a lot of people on x complaining right now, oh, Claude's getting worse. No. That's a skill issue.

02:05They're just not managing their context well. But how do you manage your context better if you start seeing worse performance with Claude? Well, what you wanna do is use their brand new slash command slash context.

02:16This is going to help you visualize what is currently in your context right now. So I'm gonna run that.

02:21And, again, slash context, I'll put down below. And you'll be able to see here now in this really nice visualization what is taking up your entire context window.

02:30So you can see here the system prompt is taking up a good amount of context. That's because I have a ton of clawed rules. All the tools we're using, the messages we've done so far, this is actually taking up the most context with 32% is the past messages.

02:44So if I wanted to free up context right now, I would just go in here and I would do slash clear, and that would clear the message history since right now it's taking up the most context. So if you start to see degraded performance in Claude, do slash context, see what's taken up all the space.

03:00Maybe it's your system message. Maybe it's your past messages in the conversation, and then you can take care of it from there.

03:06If it's the past messages, just do slash clear just like I did here, which takes me to the next big update that they released in the last few weeks, and that is slash stats. It's another brand new slash command. So one of the biggest complaints I hear about Claude as well is people don't know how many tokens they've used up in their plan.

03:23Are they getting to the end of their plan? Are they gonna have upgrade to a more expensive plan? Well, now you'll be able to see exactly how much usage you've had in your plan.

03:30If you do slash stats, it will show you all your usage for that time period. You can see all your interesting stats here, how many tokens you've used. Are you getting to your limit on tokens?

03:41Your streaks, how many sessions you've had, your active days, your peak hours. You can even tap and see the different models you've been using. As you can see, once Opus dropped, it was game over.

03:51I was using Opus tremendously more. So you can see a lot of your usage here, how many tokens you've used, and if you're getting up to your current limits and if you need to upgrade or not. Slash stats, keep tabs on that.

04:03It's also just fun to see your numbers and see how much you're using Claude and which models you're using. If you run slash stats right now, let me know down below what your usage looks like. How many days in a row have you used Claude code?

04:13I'd be really curious to see. So second ago, we were talking about resuming sessions.

04:18This is something you need to be doing often because if you're anything like me, you're also doing multiple sessions at once. I like to have multiple Claude codes open so I can work on different features at the same time. Well, one really awesome thing they just added in their recent holiday updates is the ability to actually name sessions so that when you resume old sessions like I showed you at the beginning of this video, you can see very clearly what you were doing in each session because you can give it those custom names.

04:45All you need to do to name your sessions is do slash rename and then give it a name for your session. So for this one, I was working on the Kanban board of our project management, so I'm gonna call it Kanban. And I'm gonna hit enter, and now our session is now called Kanban.

05:01So if I go to a new terminal here and I do Claude slash slash resume and hit enter, we can see our new name session Kanban that we were using just a second ago with all 246 messages.

05:14So this is really important to do. If you're doing multiple sessions, which I think you should be doing when you're waiting for one session to build out a feature, it's really good not to get distracted, not to look at TikTok. Open up another session.

05:27Start working on another feature. Make sure to give it its own name with slash rename. And, again, all these slash commands, all these updates down below in the description you don't have to forget any of these.

05:36The next update they made is great if you often make mistakes when working with Claude code. This is something I do a lot. I do a lot of experimentation.

05:44I have a change around features to see how it works, and sometimes I mess up and I give it a bad command. So for instance, in this project management tool, I wanted to experiment with the colors and notes. I made it blue.

05:54I don't like the blue. It's not really standing out against the black background. So let's rewind this.

05:59In a brand new update to Claude code they just released, you can now rewind super easily any changes you made that were mistakes. You do this by hitting escape twice.

06:08Double escape, and it goes back and it gives you these rewind options. And what I can do is I can go back and I can say rewind to the rename Kanban, which was one step before the changing of the colors of the notes.

06:21I'm gonna hit enter on that, and I'm gonna say restore the code and the conversation. So it's gonna forget any of the conversation after, and it's gonna restore all the code.

06:30And I'm gonna hit enter, and boom. Just like that, we are back. And it's at the rename Kanban, and I can hit enter on that.

06:37And we are back to that one step before changing all the colors to blue. This is amazing.

06:43If you make mistakes often, if you experiment, you should be experimenting a ton. Ever since they released this double escape feature, it's given me a lot more freedom to try new things. So double escape, an awesome way to rewind if you make any mistakes at all.

06:56The next update is a big one, and it brings back a much loved feature that's kind of been experimented with, taken out, put back in. Well, it's fully back. UltraThink is totally back into Claude code, and you need to be using it.

07:08What UltraThink is is the ability to force Claude code to work extra hard on tough problems. And I've been using this a lot, and let me tell you, you get way better results when using UltraThing. So as you can see here, I wanna add a calendar to our app.

07:22I want it to be beautiful. I want Claude to work extra hard on it. So I said, add a beautiful calendar section to our app using the design skill.

07:29Ultrathink about this. And for the record, if you don't have the design skill yet, I'll put a link down below. You need to install the Claude design skill.

07:36I'm thinking about doing an entire video dedicated to Claude skills. If you want that, let me know down the replies as well. But the key here is I said Ultrathink.

07:44And as you can see, they put it in, like, the rainbow colors. All you need to do in any of your messages is type the term Ultrathink, and it will force Claude code to think very, very hard about what it's doing and put more effort in the creation.

07:57Hitting enter on that will start the process, and it is going to make sure that this calendar looks way better than it would before. I wouldn't use UltraThink on every single prompt you do. It's gonna take a lot longer, and you're probably gonna burn a lot more tokens than you need to.

08:13But any prompts you give where you're doing a complex task or you're solving a complex bug, I would a 100% use UltraThink because you're going to get noticeably better results. This is a really good feature to use when you're kinda building out the base of your app after the first planning phase because you're gonna get a way better base to your app.

08:31Use UltraThink, again, down below in the description. The next big feature they released over these last few weeks that I absolutely love is the ability to create custom memories. So before anytime you wanted Claude to remember something, you had to go into your Claude rules and put it there.

08:47But that could be challenging at times. I don't always remember to put things into Claude rules, or Claude rules can get very crowded and disorganized if you're adding a ton in there.

08:56So what I like to do when I have small things I want Claude to remember in its process is to create custom memories for it. So for instance, I want Claude to use the design skill anytime it changes the UI, and I want it to remember that for the rest of the time that I'm working on this app.

09:14So what you wanna do to create custom memories for Claude is you're gonna do the hashtag, which changes you to memory mode. So you can say add to memory.

09:22And then we're gonna say, always use the design skill when changing the UI.

09:29And I'm gonna hit enter. And now it's going to memorize this. I wanna put it into the project memory, so you can do user memory or project memory.

09:37I want it to remember it just for this project. I'm gonna hit enter there. And now anytime we change the UI in this project, it is going to use that design skill.

09:46I actually highly recommend you copy exactly what I just did now. Go pause this video. Go into Claude code.

09:52Do hashtag, and then type in use the design skill anytime you change the UI because the design skill is so freaking good. It's gonna make your app look so much better. So custom memories, anytime you want Claude to remember something that does every single time in its process, make sure to use custom memories.

10:08I highly recommend you do the design skill memory I just did, and you can use it moving forward. This next update is a tip for the risk takers out there, but this is putting Claude into YOLO mode. This is the mode it goes into when you don't want to ask you for permission anymore.

10:24I'm gonna be completely honest with you. Claude for me hasn't made mistakes in months. And I know that might sound crazy, but when you get so used to Claude code that you know how it works, you know how it operates, you know how to talk to it, you really don't mess up with it anymore.

10:38Opus four five has gotten so smart, I don't mess up. So I like to use YOLO mode. And the way to do that is you put in this command, Claude dash dash dangerously skip permissions, And then you hit enter on that, and that's going to put Claude in YOLO mode.

10:52I'm gonna hit accept on this. I would only recommend using this if you are super used to Claude codes. If you're on the pro side of Claude code, I would do this.

11:02This saves me a ton of time. I like to give Claude long running tasks where I can hit enter and walk away, and I don't have to keep hitting accept, accept, accept over and over again. This allows me to just do that and treat Claude as an independent employee.

11:16I put the command down below. If you're Claude pro, copy that. If you're an amateur or you're a beginner, I'd say spend a month or two getting used to Opus and how it works and the right prompts to give it.

11:25And then when you're comfortable, you can use this command. I'll put that down below. Claude YOLO mode, really, really helpful for saving time.

11:31The next big update, and this is a huge one, basically adds an entire app store to Claude code, is slash plugins. If you do slash plugins, hit enter, you now have an entire, basically, store of skills and plugins and MCPs you can put right into Claude code very, very easily.

11:49I would highly recommend, if you haven't yet, go in here, do slash plug ins, and go down to front end design. This is the plug in you need if you're building any sort of app. My UIs with my apps I built are a thousand times better because of this skill.

12:04So if you haven't installed this yet, do slash plugins, go in Discover, go down to front end design, and install that.

12:11But there's so many other ones. If you use GitHub a ton, if you use any of these others, Figma, Linear, Notion, you have a ton of plug ins and MCPs. You can get installed with one click that will make it so much easier to work between all your different tools.

12:26I couldn't recommend enough. Check out the plug in store and see if you use any of these integrations inside the plug in store. At the very least, get the front end design skill installed.

12:36The last update we'll go over here is saving your prompts. If you're anything like me, you come up with an idea, you write a prompt out, and then right before you hit enter, you realize you need to do something else right before you send that prompt, and then you sit there and you hold backspace for five minutes while it deletes your prompt.

12:51You don't have to do that anymore. Say you write out a prompt. I want you to build out a calendar, and I realize I need to do something else first.

13:00You can now hit control s. That will stash your prompt. And now after I say something like add a light mode to the app and hit enter, it will take my stash prompt and put it back into the line here so I can send it after.

13:17No more sitting there and hitting back space for five minutes at a time to get rid of your prompts. Now you can just stash your prompt, go in, write whatever else you need to do, and then your stash prompt will then automatically be placed inside your command line so you can use it again.

13:33This has saved me so much time. I will put that down below as well. Control s to stash your prompt.

13:40Those were the big changes from the last month, and they're adding a ton more as we speak. They've been adding new ones every single day. I'm gonna keep an eye on them.

13:48I'm going to create more follow-up videos over the next few weeks going over all the additional changes they're coming out with. Make sure to stay on top list by subscribing, turning on notifications, and leaving a like down below.

14:00I also livestream Monday, Wednesday, Friday at 11AM Pacific Standard Time. If you wanna see me build apps out in real time with Claude Code as well. So appreciate if you watch the video.

14:10Let me know in the replies which of these updates will be the most helpful for you. I'm curious which ones you're going to use. Thank you so much for watching the channel.

14:19Every single view and like and all that means the world to me. It's an honor to be creating you educational content. I will see you in the next video.

The Hook

The bait, then the rug-pull.

Snowflakes fall across the presenter via AR overlay before the camera cuts clean to a warm studio Christmas scene and a claim that lands like a whisper from a private Slack: Anthropic has been shipping Claude Code features quietly, and if you were not on the right Twitter accounts, you missed all of them.

Frameworks

Named ideas worth stealing.

01:58model

Context Management Loop

Run /context to see what is filling the window
Identify the biggest culprit (usually messages at 32%)
Run /clear to reset messages, or trim CLAUDE.md if system prompt is the issue

Three-step mental model for diagnosing and fixing Claude hallucination caused by context overflow

Steal forAny Claude-getting-worse troubleshooting content or CLAUDE.md hygiene post

07:14concept

UltraThink Trigger Word

Adding the literal word Ultrathink to a prompt activates extended reasoning mode; renders in rainbow in the terminal; best for complex tasks and initial app scaffolding

Steal forComplex task prompts, initial app scaffolding, deep bug-dives

10:00concept